Dear developers,

We have a question concerning the storage of time series: Is it allowed to put all volumes in one data array?
In several GIfTI examples on this website each volume was put in a separate data array with it's own header. This seems a bit redundant.

Thanks,

Joost and Hester

Hester,

The convention myself and others have followed is to place each time
point in its own data array.

When compressing or uncompressing data, a contiguous piece of memory
is needed that holds the entire, uncompressed data array. If somebody
had a very dense surface with many time points and wanted to compress
it, would a contiguous piece of memory large enough be available?

Perhaps the committee needs to explicitly state how the data arrays
should be formated in each of the different file formats. I will
suggest the following:

Coordinates File
-----------------------
One DataArray with Intent=NIFTI_INTENT_POINTSET and
DataType=NIFTI_TYPE_FLOAT32. Dimensionality is 2 with the first
dimension set to the number of nodes and the second dimension set to 3.

Functional (statistic) File
---------------------------------
One or more DataArrays with Intent set to NIFTI_INTENT_NONE or one of
the statistical intents. Each DataArray has
DataType=NIFTI_TYPE_FLOAT32. Dimensionality is 1 with the first
dimension set to the number of nodes.

Label File
--------------
The file consists of a LabelTable followed by one or more DataArrays.
Each DataArray has DataType=NIFTI_TYPE_INT32. Dimensionality is 1
with the first dimensions set to the number of nodes.

RGB/RGBA File
---------------------
The file consists of one or more DataArrays with Intent set to
NIFTI_INTENT_RGB_VECTOR or NIFTI_INTENT_RGBA_VECTOR. Each DataArray
has DataType=NIFTI_TYPE_UNIT8. Dimensionality is 2 with the the first
dimension set to the number of nodes and the second dimension set to 3
for RGB data or 4 for RGBA data.

Shape File
---------------
One or more DataArrays with Intent set to NIFTI_INTENT_SHAPE. Each
DataArray has DataType=NIFTI_TYPE_FLOAT32. Dimensionality is 1 with
the first dimension set to the number of nodes.

Surface File
----------------
A surface file contains two DataArrays. The first DataArray has
Intent=NIFTI_INTENT_POINTSET and DataType=NIFTI_TYPE_FLOAT32.
Dimensionality is 2 with the first dimension set to the number of
nodes and the second dimension set to 3. The second DataArray has
Intent=NIFTI_INTENT_TRIANGLE and DataType=NIFTI_TYPE_INT32.
Dimensionality is 2 with the first dimension set to the number of
triangles and the second dimension set to 3.

Time Series File
----------------------
One or more DataArrays with Intent set to NIFTI_INTENT_TIME_SERIES.
Each DataArray has DataType=NIFTI_TYPE_FLOAT32. Dimensionality is 1
with the first dimension set to the number of nodes.

Topology File
------------------
One DataArray with Intent=NIFTI_INTENT_TRIANGLE and
DataType=NIFTI_TYPE_INT32. Dimensionality is 2 with the first
dimension set to the number of triangles and the second dimension set
to 3.


All of the multidimensional DataArrays are assumed that
ArrayIndexingOrder is RowMajorOrder. Do we really need to allow
ColumnMajorOrder?


John Harwell


>

Dear John,

Thank you for your quick reply. We have followed the convention that you and others used in our implementation.
Concerning the other points, maybe this can be discussed during the upcoming phone conference?

Cheers,
Hester

Hi
So, has this issue been discussed about ?
Could the suggestion posted above be considered a standard or a best-practice ?
Thanks